AirwaySim
Online Airline Management Simulation
Login
Username
Password
 
or login using:
 
My Account
Username:
E-mail:
Edit account
» Achievements
» Logout
Game Credits
Credit balance: 0 Cr
Buy credits
» Credit history
» Credits FAQ

Author Topic: Demand modeling questions  (Read 7124 times)

Offline alexgv1

  • Members
  • Posts: 2273
Re: Demand modeling questions
« Reply #20 on: March 09, 2013, 10:59:52 PM »
Since the PAX allocation system does not make a difference between huge amounts of seasonal leasure PAX and the average traffic (business)
PAX, it should be very difficult to get a reasonable all-year-average demand between many destinations.
Maybe the new city based demand model may bring the answer as regions can be differentiated as business/family or mainly leisure areas.

It would be nice for the pax to be split between business (frequency/service sensitive) and leisure (price sensitive ad more seasonal). This was an old idea but I'm not sure how necessary it is as this was from before the new pax allocation. Sami says seasonality wouldn't be modelled so some average will be taken I guess. City based demand our answer to everything haha.

(Although actually the game is great as it is and I am really enjoying playing it again after an absence of 5 or so months)
CEO of South Where Airlines (SWA|WH)

Offline Seattle

  • Members
  • Posts: 2791
Re: Demand modeling questions
« Reply #21 on: March 10, 2013, 12:22:30 AM »
Seattle and Algeria *shudder* I still have fond memories of being soundly beaten, and that was back in beta...  ::)

Hehe ;D

Though, after around two years of absence, I'm ashamed to say, I have not yet crushed the competition at ALG.  :-[
Founder of the Star Alliance!

Offline Andre

  • Members
  • Posts: 1090
Re: Demand modeling questions
« Reply #22 on: March 10, 2013, 05:46:17 AM »
Since the PAX allocation system does not make a difference between huge amounts of seasonal leasure PAX and the average traffic (business)
PAX, it should be very difficult to get a reasonable all-year-average demand between many destinations.
Maybe the new city based demand model may bring the answer as regions can be differentiated as business/family or mainly leisure areas.

Stop over complicating things, exchlbg. Everything does not have to be super specific, and calculating a reasonable all-year average is simple math. There's no need for a perfect ratio between business or leisure travel.

Budjet

  • Former member
Re: Demand modeling questions
« Reply #23 on: March 10, 2013, 08:37:06 AM »
I think what also needs to be taken into consideration which may be complicated however is connecting passengers if there isn't a direct route... for example if there is no flight between Stansted and Prague, however there is a flight from Stansted to Amsterdam and a flight from Amsterdam and Prague, then more people will take this flight, a real life example would be that I fly between Bremen and Riga however we stopped flying to Tallinn for the winter, therefore we get many Estonians on our Riga flight who then take connecting flight to Riga and when we operate Porto we get people who are flying there to catch a long haul flight to South America.

Away that this could be implemented is that if there is no flight between point A and point B, but a flight from A to C which then has a flight from C to B, then there demand between the routes is higher unless someone opens a direct flight between the two routes. For example:

50 people from Amsterdam want to get to Istanbul, 80% of them are willing to catch a connecting flight to London leaving a demand for the connection as 40.
Meaning the demand for the Amsterdam to London is 40 pax higher, then the demand between London and Istanbul is also 40 pax higher.
However later in the year your airline opens a direct flight between Amsterdam and Istanbul meaning the demand for the Amsterdam to London and London to Istanbul flight drops 40 pax... this is just an example, obviously it would have to take into account how convenient the connection is, for example no one would buy a ticket from Amsterdam to New York from New York to Istanbul, you'd have to find a way for the system to recognize the quickest connections to only increase the demand on those routes.... man this is getting complicated, can someone think of another way?
« Last Edit: March 10, 2013, 08:46:11 AM by Jules »

exchlbg

  • Former member
Re: Demand modeling questions
« Reply #24 on: March 10, 2013, 10:02:52 AM »
Jules, that are great ideas, but do you really think, those wernīt had already? Discussions about how to bring connectivity into game without a complicating things would fill books.
Normally itīs me whoīs talking against changing demand model because of discrepancies between sim and RL, I did so earlier in this thread.
The closer other players want this game tied to RL numbers, the more complicated everything gets, thatīs for sure.
If it was me to decide I would leave it mostly how it is, demands calculated on a fixed formula without too much exceptions.
We have to keep the average player in mind, who does not have the time to do special researches about possible cultural ties between two nations in any era.
Likewise I would ask all other players to stop complaining about discrepancies in PAX numbers to here or there. We are simulating airlines in simulated worlds, not replaying todays schedules all the time. Fot every demand that seems to be too small in here thereīs dozen of others which donīt apper in RL. Overall traffic number are higher here, anyway.
« Last Edit: March 10, 2013, 10:06:48 AM by exchlbg »

Offline Sami

  • Administrator
  • Members
  • Posts: 16486
    • AirwaySim - Are you the next Richard Branson?
Re: Demand modeling questions
« Reply #25 on: March 13, 2013, 02:54:44 PM »
These country relations can be easily modeled even with the current systems if we'd have the data for it. But for most of the world, we don't as it's pure manual work in forming the country-pair relations (over 50 year span in history) and using judgement to determine the level of demand... Volunteers are welcome as always.

In regards to my earlier message, I would be happy if someone (1-3 guys) would be willing to work on this.

Some info and instructions:

Basically I am looking for a table (excel for example) that has a list of countries (according to their AWS name/id) and a "relations index" towards other countries. As discussed here, it would be a simple numerical index from 0.0 to 3.0 that measures the travel relationship between the two countries.

- A high relationship is usually because of historical factors; the country was a former colony of the parent, or there have been a lots of immigrants from there. Or in modern times the country supplies a lot of immigrant workers who travel back and forth. For example factor between France and Algeria could be something around 3.0 as they are historically closely linked.

- For poor relationships examples would be mostly political, for example North Koreans are not allowed to fly directly to South Korea. So that factor would be zero.

- Standard value is 1, and if relationship between countries is "normal", then there's no need to change anything. For example Belgium and Australia have no special ties or links, so they'd have a default factor of 1.0. Most of the country pairs are like this.

The number, the factor (0-3) is a number that is factored on top of the standard demand by multiplying it (let's say if France-Algeria demand would be 100, with this factor it would be 300).

Any temporary events like wars or such are not counted here. Also if countries are just located close together (and hence have higher demand) does not apply to these factors.

What the actual number between each country is, is left upon the discretion of the data collector. Most of all I would ask to use common sense and knowledge + sources, and decide a number that is relevant judging the info (ie. 3 = very close ties, 1 = normal relations, 0 = all travel disallowed). The number can have decimals so 2.5 is allowed for example. Do not look at the current demand of AWS games as it may be misleading.


If you are interested, PM me and I can drop you the excel file. Of course all country relations (~200 countries) do not have to be filled, so it's not an immense work I reckon. There are some entries in the database already on this..

(I think I got a PM about this from someone earlier, but may have accidentially deleted it)

Offline Julien

  • Members
  • Posts: 574
Re: Demand modeling questions
« Reply #26 on: March 18, 2013, 10:44:24 AM »
Hello,

I already made an Excel file, I asked you before in this post where to send it ?
I think you deleted my PM also.

I have a question : you are talking about countries. Can we consider states or cities ?
Many links above can't work considering only countries.

If anyone can help me by adding datas... here are the rules :
FEEL FREE TO FILL IN WITH RELEVANT INFORMATION
Then we will proceed with exact demand (0 to 3)

1/ Demand influenced by political reasons :
*Cuba to/from US = 0 except Miami
*North to/from South Corea = 0
*Armenia to/from Turkia = very low
*Israel to surrounding Arab countries = very low
*Pakistan to/from India = quite low
*Lhasa in Tibet can't be reached from outside China
Etc...

2/ Demand influenced by historical reasons :
A/ Former parents to colonies :
*France / UK / USA / SPAIN to current overseas territories (only Ceuta and Mellila for Spain)
*French / British / American overseas territories between themselves if geographically close
*France to former colonies : West Africa / Vietnam / North Africa / Djibouti / Lebanon
*Spain to former colonies : South America except Brazil / Fillipines / some Carribean destinations
*Portugal to former colonies : Angola, Mozambique, Brazil, Cabo Verde
*Belgium to former colonies : Congo
*The Netherlands to former colonies : Indonesia / South Africa / Suriname / some Carribean routes
*Russia to ex soviet bloc
*Denmark to Scandinavia
*Istanbul has both strong natural demand to/from Europe and Asia

3/ Demand influenced by cultural reasons :
A/ Strong demand between same language countries (considered as very old immigration):
FRENCH
*Major cities of major countries where french is official language : France - Canada - Switzerland - Belgium - - North Africa countries - West Africa countries - some carribean + overseas
ENGLISH
*Major cities of major countries where English is official language : USA - Canada - UK - Ireland - Kenya - Singapore - India - Pakistan - South Africa - some carribean + overseas
SPANISH
*Major cities of major countries where Spanish is official language : Spain - Mexico - Argentina - Colombia - Peru - Dominican Republic - Venezuela - Ecuador - some carribean + overseas
MALAYSIAN
*Major cities of Sumatra (Indonesia) - Malaysia - Singapore and Brunei linked together
ARABIC
*Major cities of North Africa - Middle East
RUSSIAN
*Taking back datas from Ex-Colonies + Mongolia (Ulan Bator)
GERMAN
*Major cities + secondary cities (as countries are all packed together) of countries speaking German : Germany, Luxemburg, Austria, Switzerland
CHINESE
*China linked with Taiwan (now opening, see China Airlines B747 daily at PVG), Malaysia and Singapore
PERSIAN
*Iran, Afghanistan and Tajikistan

B/ Strong demand between countries of a same "cultural bloc"
HINDI/HINDU BLOC
*Nepal/Bhutan/India/Sri Lanka and Bangladesh (major cities)
SCANDINAVIAN BLOC
*Denmark, Iceland, Norway, Finland, Sweden
CENTRAL ASIA BLOC
*Azerbaidjan, Afganistan, Tadjikistan etc...
EASTERN EUROPE BLOC
*Eastern Europe countries formerly linked with Russia still remain very linked together
IBERIC PENINSULA
*Portugal >< Spain
ASIAN PENINSULA
*Thailand, Malaysia, Laos, Birmania, Vietnam

4/ Demand influenced by geographic reasons : (Regional Effect - same cultural links affects travel, can be language, position...)
*Demand between countries and overseas territories of Caribean
*Demand between countries and overseas territories of Caribean ++ when same language
*Demand between countries and overseas territories of Pacific Ocean
*Demand between countries and overseas territories of Pacific Ocean ++ when same language
*Demand between countries and overseas territories of Indian Ocean
*Demand between countries and overseas territories of Indian Ocean ++ when same language
*Demand between countries and overseas territories of Caribean and the regional capital considered Miami
*Demand between countries and overseas territories of Pacific Ocean and their regional capitals considered Auckland and Sydney

5/ Demand influenced by recent immigration (regular demand from families / workers etc abroad)
*Turkia >< Germany
*China >< Major East african cities
*China >< Malaysia + Singapore
*California >< Latin America
*Miami >< Latin America
*Major US West cost cities to major Asian cities
*Major + secondaries North African Cities with major french cities (already in historical but here too as second wave of immigration in 80s 90s)
*Pakistan >< UK
*Albania >< Italia



EXCEPTIONS :
Actually you had maybe the idea to include those datas but there are no longer available :
*Taiwan/China = 0 is no longer available as Taipei is now directly linked with major chinese cities
*Filipines (former Spanish colonies) is not linked anymore to Spain for decades as demand is not strong enough anymore


I am now filling the demand quantity from 0 to 1 if negative and from 1 to 3 if positive by searching real traffic, information and sources.
SEASONAL ROUTES NO COMPILED

Thanx.
« Last Edit: March 18, 2013, 12:04:58 PM by Julien »

Offline Seattle

  • Members
  • Posts: 2791
Re: Demand modeling questions
« Reply #27 on: March 22, 2013, 06:36:23 AM »
Sami, I pmed you a couple days ago as well. I'd be happy to work with Julien to come up with the country relations data.

Julien is that the criteria you came up with or what Sami did? Anyway, I'm fairly knowledgable about such historical connections and looking up current flights is fairly easy.
Founder of the Star Alliance!

Offline marchayes1986

  • Members
  • Posts: 5
Re: Demand modeling questions
« Reply #28 on: March 22, 2013, 01:20:12 PM »
Just to join in the discussion about demand and transfer, I wondered if the index system which Sami has proposed could be extended to certain or particular airports.

There is a lot of discussion in the forum about the complexities of modelling transfer traffic and how airport-to-airport demand reflects this.  If I take for example my local airport, Durham Tees Valley (EGNV/MME), you can now only fly to either Aberdeen or Amsterdam and the latter route is served by KLM three times daily with a F70.  The recession has clearly affected the local area and the selection of flights offered from the airport, compared to the success enjoyed several years ago.

My point is that to get anywhere international from MME you need to fly via AMS and change with KLM or another airline to your onward destination (or drive an hour north or south and fly from Newcastle or Leeds Bradford direct if they offer such a service).  I'd be pretty sure this situation is representative of most airports with connections to hub airports, unless a direct flight is offered and the service/flight times appeal to passengers. 

When this game world started the demand to certain hub airports such as LHR, CDG and AMS was already very strong which emulates the real world to an extent despite those airports not providing a 'hub and spoke' network immediately.  Could demand from smaller airports to major airports be more dynamic in that the demand reflects the amount of potential connections offered as well as any inherent demand on the route? 

For example (and these figures are arbitrary)...

If Airport A has flights to 5 or less destinations:
the demand to airports which have over 200 destinations is multiplied by 3.
the demand to airports which have over 100 destinations is multiplied by 2.5
the demand to airports which have over 50 destinations is multiplied by 2.

If Airport B has flights to 20 or less destinations:
the demand to airports which have over 200 destinations is multiplied by 2
the demand to airports which have over 100 destinations is multiplied by 1.4
the demand to airports which have over 50 destinations is multiplied by 1.05

The frequency, route image and aircraft type then play their usual part in simulating how many passengers fly with your airline.  If an airport with few destinations suddenly sees itself with new flights to an increased number of destinations, this would then affect the demand on the route.

Perhaps this suggestion may raise issues I have not predicted for which I apologise, but I merely submit this suggestion as a potential solution to the connections debate and as a way of allowing players to establish non-traditional airports as hubs in this virtual world.

Thanks

Offline Seattle

  • Members
  • Posts: 2791
Re: Demand modeling questions
« Reply #29 on: March 22, 2013, 07:44:01 PM »
I don't know if you know, but AWS will be switching to a city based demand system, where airports simply serve sets of population (with a few other factors, such as an area's business/industrial output and tourism). How exactly this system will work is still being ironed out, but like you suggested, flights to certain airports will affect those to other ones, as each airport has a catchment area (and can "steal" passengers from neighboring catchment areas, regardless of an airport being there).

What you're also suggesting is connecting passengers and I am not sure what's going on in terms of development, but I'm pretty sure Sami is focused on getting the new demand system in place before touching that.
Founder of the Star Alliance!

exchlbg

  • Former member
Re: Demand modeling questions
« Reply #30 on: March 22, 2013, 08:45:53 PM »
To marchayes: since the base data for existant demand model relate to reality, they already reflect the fact of transfer PAX. Otherwise the demand from small airports to hubs would be much smaller.All tries to model connections on top of these data would lead to an widely exaggerated air traffic.Worldwide numbers already are much higher as in reality.If Sami would imply connecting traffic, then hub traffic numbers generally will be much,much smaller before users are starting their conections there.

Offline marchayes1986

  • Members
  • Posts: 5
Re: Demand modeling questions
« Reply #31 on: March 23, 2013, 03:42:56 AM »
Thanks for the responses. 

I realise that the issue of connecting passengers poses considerable issues.  My idea isn't necessarily to create connecting passengers within the sim, but to reflect the possibility for smaller airports with low demand to most other airports to experience increased demand to airports which offer good connectivity.

My proposal/suggestion was to replicate the demand from smaller airports to larger 'hub' airports based on the connectivity of the hub airport.  Exchlbg suggests that airport demands reflect reality, but surely this then restricts the growth of virtual carriers within the Airwaysim world.  If my airline is based in Rome or Madrid and I offer worldwide connections, shouldn't  the demand from a feeder airport to my hub be increased because of the connectivity offered and not just be based on real world demand, which may in fact be minimal?

exchlbg

  • Former member
Re: Demand modeling questions
« Reply #32 on: March 23, 2013, 10:07:15 AM »
Yes.With city-based demand we hope these issues will be solved, itīs a new demand model in the making.You can learn about it in the announcement forums. Thatīs why changes to current system are unlikely to happen.

hotboidavz

  • Former member
Re: Demand modeling questions
« Reply #33 on: April 11, 2013, 03:17:10 AM »
Ok, so I did not read all the post on this topic but I can say that the demand modeling between the North America-Caribbean & Europe-Caribbean is completely waaaaaay off and inaccurate. It really needs to be looked at and fixed so it reflects the real world model, the caribbean is a big market and it is poorly represented in this game.

brique

  • Former member
Re: Demand modeling questions
« Reply #34 on: April 11, 2013, 03:34:43 AM »
Ok, so I did not read all the post on this topic but I can say that the demand modeling between the North America-Caribbean & Europe-Caribbean is completely waaaaaay off and inaccurate. It really needs to be looked at and fixed so it reflects the real world model, the caribbean is a big market and it is poorly represented in this game.

..and a lot of that demand is seasonal : its holidaymakers, a fair chunk of whom fly charter on package deals. Currently, the game doesnt model such seasonal charter traffic as it would complicate matters enormously, both for the game itself and for players : it would require spare planes and the player opening new routes at start of season, then closing them at end. This would create havoc for RI and staffing, just to think of two areas not best adapted to seasonal work in the game. In the rw, charter traffic can be moved around, the planes get redeployed to where its high season elsewhere, temporary staff can be employed, pax levels (revenue, in other words) are guaranteed by the holiday companies by pre-booking capacity. None of that can be easily replicated.

The game does its best to replicate traffic between points over the year without too many major upheavals : it tries to reflect the general popularity of those routes : its not perfect, but buy Sami a roomful of Crays and a few hundred code monkeys, and he probably could build such a model... but, I fear, not this year.

Offline TFC1

  • Members
  • Posts: 787
Re: Demand modeling questions
« Reply #35 on: April 11, 2013, 06:28:48 AM »
To add a bit of info for Scandinavia, demand from Norway to Spain and the Canary Islands is strong. Typical destinations are Alicante, Malaga and Barcelona due to a large number of pensioners residing in Spain. I would also include oil-related demand between Norway and certain destinations in the US and UK, typically Aberdeen and Houston.

In addition, there are cultural ties between Scandinavia and some states in the US due to immigration. Minnesota, Wisconsin, North and South Dakota have a relatively large population of Scandinavian heritage, and that generates some demand.


Offline Andre

  • Members
  • Posts: 1090
Re: Demand modeling questions
« Reply #36 on: April 11, 2013, 12:14:25 PM »
To add a bit of info for Scandinavia, demand from Norway to Spain and the Canary Islands is strong. Typical destinations are Alicante, Malaga and Barcelona due to a large number of pensioners residing in Spain. I would also include oil-related demand between Norway and certain destinations in the US and UK, typically Aberdeen and Houston.

In addition, there are cultural ties between Scandinavia and some states in the US due to immigration. Minnesota, Wisconsin, North and South Dakota have a relatively large population of Scandinavian heritage, and that generates some demand.



I can vouch for that. It seems that the UK - Spain traffic is boosted, and the Norway - Spain traffic should be boosted as well. Norway - US traffic has to increase as well, right now it's very very limited. While in real life, there are multiple departures to the US from Oslo daily. At least enough to fill up an A330/A340, a B757 and a B787. The rest fly via Stockholm/Copenhagen/Frankfurt/Amsterdam. Either way, the Norway - US traffic should be boosted by a 3x factor compared to what it is now. Oslo - Bangkok should be increased too.
« Last Edit: April 11, 2013, 12:16:34 PM by [SC] Andre »

Offline LemonButt

  • Members
  • Posts: 1895
Re: Demand modeling questions
« Reply #37 on: April 11, 2013, 01:11:48 PM »
I'm pretty sure the traffic is there for most of these routes, but all the pax are funnelled through a major international gateway hub like JFK or ORD, which is why their demand is so much higher than an airport like Appleton or Fargo.

exchlbg

  • Former member
Re: Demand modeling questions
« Reply #38 on: April 11, 2013, 02:39:32 PM »
If you boost traffic on one side, you will have to take that boost away from other relations, like all demands between small airports.I never hear a wish to set every combination that isnīt served by a real-life airline to zero. All you wish wish for is boost, especially for that part of the world you are just sitting in.
You canīt hear anything about all other markets across the world.
Sami possibly canīt control every airport timetable of the world and design demands as-is, especially because we are not playing just traffic of today. What about 1952 dates or 2020 ?
If I look at my home airport, game demands never meet actual numbers, in any age.
Why not wait for new numbers of city-based demand ? Maybe they differ a lot. And as long we are not able to build connecting traffic, we are not able to rebuild "realistic" traffic.
Since every model is based on worldwide traffic numbers, a wish for a boost should always include a tip, where other demands are too high and can be scratched, to keep overall balance.
« Last Edit: April 11, 2013, 05:18:43 PM by exchlbg »

hotboidavz

  • Former member
Re: Demand modeling questions
« Reply #39 on: April 12, 2013, 02:47:59 AM »
..and a lot of that demand is seasonal : its holidaymakers, a fair chunk of whom fly charter on package deals. Currently, the game doesnt model such seasonal charter traffic as it would complicate matters enormously, both for the game itself and for players : it would require spare planes and the player opening new routes at start of season, then closing them at end. This would create havoc for RI and staffing, just to think of two areas not best adapted to seasonal work in the game. In the rw, charter traffic can be moved around, the planes get redeployed to where its high season elsewhere, temporary staff can be employed, pax levels (revenue, in other words) are guaranteed by the holiday companies by pre-booking capacity. None of that can be easily replicated.

The game does its best to replicate traffic between points over the year without too many major upheavals : it tries to reflect the general popularity of those routes : its not perfect, but buy Sami a roomful of Crays and a few hundred code monkeys, and he probably could build such a model... but, I fear, not this year.


Actually its NOT Just Seasonal. Maybe if you did your research you would be a bit more informed. The demand is DAILY and with more of a demand on the weekends (Friday-Sunday) for obvious reasons. And this daily demand is for over 25 destinations and its from the major hub cities on the East Coast of the U.S, Canada, the UK, France, Amsterdam, Germany and the larger market in South America. Yes there's seasonal traffic and charter but that only accounts for a small % of the Yearly traffic coming to the caribbean.

 

WARNING! This website is not compatible with the old version of Internet Explorer you are using.

If you are using the latest version please turn OFF the compatibility mode.